Identifying Alcohol Use Disorder With Resting State Functional Magnetic Resonance Imaging Data: A Comparison Among...

Area of Science:

• Neuroscience
• Computational Psychiatry

Background:

• Alcohol use disorder (AUD) presents significant societal and health challenges, with brain changes often difficult to assess.
• Comorbid substance use, like nicotine, complicates the study of AUD's neurological effects.
• Machine learning (ML) offers promising tools for analyzing complex neurological data and detecting AUD.

Purpose of the Study:

• To evaluate the performance of various ML classifiers in detecting AUD.
• To identify key brain functional connectivity patterns associated with AUD.
• To assess AUD detection in the absence of other substance use, specifically nicotine.

Main Methods:

• Utilized resting-state functional network connectivity (rsFNC) data from 51 AUD subjects and 51 controls, all free of other substance use.
• Applied independent component analysis to derive rsFNC estimates and employed a random forest approach for feature selection.
• Evaluated 10 ML classifiers, including neural networks, logistic regression, and support vector machines, using classification performance metrics.

Main Results:

• A neural network classifier achieved the highest performance with an Area Under the Curve (AUC) of 0.79.
• Logistic regression, nearest neighbor, and support vector machine classifiers also demonstrated strong performance.
• Key features for classification involved functional connections within visual, sensorimotor, executive control, reward, and salience networks.

Conclusions:

• Machine learning classifiers can effectively identify alcohol use disorder (AUD) using rsFNC data.
• The absence of nicotine comorbidity did not hinder the accurate detection of AUD.
• This research highlights the potential of ML in diagnosing AUD based on distinct neural connectivity patterns.